Arena Series B: $200M Raise at $3.1B Rests on Usage Revenue

What happened

Arena, the AI evaluation company best known for the crowdsourced model leaderboard that AI labs routinely cite when they ship new systems, has closed a $200 million Series B at a $3.1 billion valuation. 14 Lightspeed Venture Partners and Khosla Ventures co-led the round. Salesforce Ventures, 01 Advisors, Dell Technologies Capital, Endeavor Catalyst, Andreessen Horowitz, Felicis and others also participated. 14 The company says it will use the money to expand evaluation of AI models and, increasingly, autonomous agents. That focus is framed around the new risks agents introduce. 14

The round comes roughly nine months after a $150 million Series A in January at a $1.7 billion valuation. 34 Counting a previously reported $100 million seed, Arena has now raised about $450 million in total. 4 That figure fits with earlier reporting that the company had raised $250 million before this round. 23

The revenue story

The headline number behind the valuation is revenue. Arena says it now exceeds $100 million in annualized revenue. 1 The trajectory is what caught investors' attention. Arena's paid AI Evaluations product launched in September 2025 and sells deeper performance analytics drawn from its user community. 3 Annualized revenue stood at about $30 million at the time of the Series A. 23 By late June, the company disclosed it had crossed $100 million, roughly eight months after the commercial launch. 2

Venture-focused coverage treated this as one of the year's fastest monetization ramps among AI-adjacent businesses. One outlet argued that tripling revenue in about five months, without new capital, made the January valuation look conservative in hindsight. 2 The same analysis cast Arena as rare "AI infrastructure for AI itself." In that view, the free public leaderboard acts as the moat that feeds the paid product. 2

The asterisk: this isn't classic ARR

The sources diverge on how much weight the $100 million figure should carry. Some coverage labeled it ARR without qualification. 2 But CEO Anastasios Angelopoulos himself said, in comments reported via TechCrunch, that the revenue is consumption-based. In his words, it is not annualized recurring revenue in the classical sense. 3 A separate analysis of the Series B made a related point. The number is company-reported and should not be treated as audited ARR. 1

The distinction matters. Classic SaaS ARR implies contracted, recurring subscriptions with relatively predictable renewals. Consumption revenue scales with how much customers use the service in a given period. Annualizing a strong recent stretch can overstate what the business will reliably earn over a full year. For an evaluation platform, usage is likely to rise and fall with model release cycles. Labs pay for intensive testing around launches, and spending may ease between them. That reading is an inference, but it is the kind of volatility consumption models tend to carry.

Reading the valuation

On the reported numbers, the new valuation works out to roughly 31 times annualized revenue. The step-up from $1.7 billion to $3.1 billion in under a year is steep but not unusual for AI companies posting this kind of growth. The investor list leans toward strategic interest. Salesforce Ventures and Dell Technologies Capital sit alongside the lead financial firms. 14 That suggests enterprises see evaluation as a necessary layer as they deploy models and agents in production.

The agent angle may be the stronger long-term thesis. Agents that take actions, rather than just generate text, raise harder questions about reliability and alignment. Arena's pitch is that independent evaluation becomes more valuable as those risks grow. 14 A company that started as a UC Berkeley research project is now being priced like a growth-stage software firm on that premise. 1

The take

Arena's growth is real by any reasonable standard. Going from a research leaderboard to a nine-figure revenue run rate in under a year is notable. Still, the "$100M ARR" framing that circulated in some coverage overstates the quality of that revenue. The company's own CEO has drawn the line, and it is worth taking him at his word. 3 Investors appear to be betting that usage-driven revenue will settle into something steadier as evaluation becomes standard practice for model and agent deployment.

There is a structural tension underneath the bet. Arena's influence rests on being a neutral public scoreboard. Its revenue increasingly comes from the very labs whose models it ranks. That combination is a strong business position, but it puts a premium on maintaining trust. The questions to watch are whether consumption revenue holds between major model launches, and whether Arena's credibility as an arbiter holds up as its commercial ties deepen.